The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 7649773293 ("cxgb3/l2t: Fix undefined behaviour")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Acked-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/20200226222722.GA18020@embeddedor
Currently when marking a block, we use spinand_erase_op() to erase
the block before writing the marker to the OOB area. Doing so without
waiting for the operation to finish can lead to the marking failing
silently and no bad block marker being written to the flash.
In fact we don't need to do an erase at all before writing the BBM.
The ECC is disabled for raw accesses to the OOB data and we don't
need to work around any issues with chips reporting ECC errors as it
is known to be the case for raw NAND.
Fixes: 7529df4652 ("mtd: nand: Add core infrastructure to support SPI NANDs")
Cc: stable@vger.kernel.org
Signed-off-by: Frieder Schrempf <frieder.schrempf@kontron.de>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/20200218100432.32433-4-frieder.schrempf@kontron.de
For reading and writing the bad block markers, spinand->oobbuf is
currently used as a buffer for the marker bytes. During the
underlying read and write operations to actually get/set the content
of the OOB area, the content of spinand->oobbuf is reused and changed
by accessing it through spinand->oobbuf and/or spinand->databuf.
This is a flaw in the original design of the SPI NAND core and at the
latest from 13c15e07ee ("mtd: spinand: Handle the case where
PROGRAM LOAD does not reset the cache") on, it results in not having
the bad block marker written at all, as the spinand->oobbuf is
cleared to 0xff after setting the marker bytes to zero.
To fix it, we now just store the two bytes for the marker on the
stack and let the read/write operations copy it from/to the page
buffer later.
Fixes: 7529df4652 ("mtd: nand: Add core infrastructure to support SPI NANDs")
Cc: stable@vger.kernel.org
Signed-off-by: Frieder Schrempf <frieder.schrempf@kontron.de>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/20200218100432.32433-2-frieder.schrempf@kontron.de
Macronix NANDs support randomizer operation for user data scrambled,
which can be enabled with a SET_FEATURE.
User data written to the NAND device without randomizer is still readable
after randomizer function enabled.
The penalty of randomizer are subpage accesses prohibited and more time
period is needed in program operation and entering deep power-down mode.
i.e., tPROG 300us to 340us(randomizer enabled)
For more high-reliability concern, if subpage write not available with
hardware ECC and then to enable randomizer is recommended by default.
Driver checks byte 167 of Vendor Blocks in ONFI parameter page table
to see if this high-reliability function is supported. By adding a new
specific DT property in children nodes to enable randomizer function.
Signed-off-by: Mason Yang <masonccyang@mxic.com.tw>
Reviewed-by: Miquel Raynal <miquel.raynal@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/1581922600-25461-2-git-send-email-masonccyang@mxic.com.tw
In order to be merged with "gpio-nand", the driver must support custom
(non-GPIO) I/O accessors.
Allow platforms to omit data GPIO port as well as NWE pin info from
device setup. For the driver to still work on such platform, custom
I/O accessors as well as a custom probe function which initialises the
driver private structure with those accessors must be added to the
driver.
Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/20200212003929.6682-14-jmkrzyszt@gmail.com
Instead of collecting partitions in a flat list, create a hierarchy
within the mtd_info structure: use a partitions list to keep track of
the partitions of an MTD device (which might be itself a partition of
another MTD device), a pointer to the parent device (NULL when the MTD
device is the root one, not a partition).
By also saving directly in mtd_info the offset of the partition, we
can get rid of the mtd_part structure.
While at it, be consistent in the naming of the mtd_info structures to
ease the understanding of the new hierarchy: these structures are
usually called 'mtd', unless there are multiple instances of the same
structure. In this case, there is usually a parent/child bound so we
will call them 'parent' and 'child'.
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/20200114090952.11232-1-miquel.raynal@bootlin.com
Currently there are 3 different variants of read_id implementation:
1. opcode only. Found in GD5FxGQ4xF.
2. opcode + 1 addr byte. Found in GD5GxGQ4xA/E
3. opcode + 1 dummy byte. Found in other currently supported chips.
Original implementation was for variant 1 and let detect function
of chips with variant 2 and 3 to ignore the first byte. This isn't
robust:
1. For chips of variant 2, if SPI master doesn't keep MOSI low
during read, chip will get a random id offset, and the entire id
buffer will shift by that offset, causing detect failure.
2. For chips of variant 1, if it happens to get a devid that equals
to manufacture id of variant 2 or 3 chips, it'll get incorrectly
detected.
This patch reworks detect procedure to address problems above. New
logic do detection for all variants separatedly, in 1-2-3 order.
Since all current detect methods do exactly the same id matching
procedure, unify them into core.c and remove detect method from
manufacture_ops.
Tested on GD5F1GQ4UAYIG and W25N01GVZEIG.
Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/20200208074439.146296-1-gch981213@gmail.com
If the write protect signal from this IP is connected to the NAND
device, this IP can handle the WP# pin via the WRITE_PROTECT
register.
The Denali NAND Flash Memory Controller User's Guide describes
this register like follows:
When the controller is in reset, the WP# pin is always asserted
to the device. Once the reset is removed, the WP# is de-asserted.
The software will then have to come and program this bit to
assert/de-assert the same.
1 - Write protect de-assert
0 - Write protect assert
The default value is 1, so the write protect is de-asserted after
the reset is removed. The driver can write to the device unless
someone has explicitly cleared register before booting the kernel.
The boot ROM of some UniPhier SoCs (LD4, Pro4, sLD8, Pro5) is the
case; the boot ROM clears the WRITE_PROTECT register when the system
is booting from the NAND device, so the NAND device becomes read-only.
Set it to 1 in the driver in order to allow the write access to the
device.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/20200127123934.11847-1-yamada.masahiro@socionext.com
The used way to compute jiffies timeout brokes when
jiffie difference is 1.
Assume that nand_soft_waitrdy is called with timeout_ms==1.
Jiffies are 1000 for example (assume something more like 1000.99
- just before incrementing to 1001).
We compute timeout_ms = 1000+msecs_to_jiffies(1) = 1001.
nand_read_data_op is called for the first time and returns 0.
During the call jiffies changes to 1001 thus "while loop" ends
here (wrongly). Notice that routine was called with expected timeout
1ms but actual timeout used was something between 0...1ms.
Fixes STM32MP1 FMC2 NAND controller which sometimes failed
exactly in this way.
Signed-off-by: Martin Devera <devik@eaxlabs.cz>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/20200116135431.17480-1-devik@eaxlabs.cz
The Hyperbus core expects that HyperFlash is always directly mapped for
both read and write, but in reality this may not always be the case, e.g.
Renesas RPC-IF has read only direct mapping. Move the code setting up the
direct mapping from the Hyperbus core to thh TI AM554 HBMC driver.
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
In case the compatible "cypress,hyperflash" is not given
output a proper error message.
Signed-off-by: Dirk Behme <dirk.behme@de.bosch.com>
Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
- Don't use `tmp` for two purposes (return value, loop counter).
Instead, use `i` for the loop counter, and `ret` for the return value.
- Don't use tabs between type and name in variable declarations,
for consistency with other functions in spi-nor.c.
- Rewrite nested `if`s as `if (a && b)`.
- Remove `info` variable, and use spi_nor_ids[i] directly.
Signed-off-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
[tudor.ambarus@microchip.com: change i's type from int to unsigned int,
reorder local variables]
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
spi_nor_spimem_xfer_data() being a helper function for the data reads/
writes contains 3 fragments that depend on the data direction; and I'm
going to add another one to call the SPI dirmap API...
I think this function should be split so that the common fragments are
put into 2 functions, spi_nor_spimem_bounce() and spi_nor_spimem_exec_op()
called from spi_nor_spimem_{read|write}_data(), and the data direction
dependent bits moved back into those read/write functions -- that way we
would be able to avoid *goto*s otherwise needed in the next patch adding
the SPI dirmap support...
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
The shift variable of SR_BP is conclusive because the first bit of SR_BP
is fixed on all known flashes. Replace ffs operation with SR_BP_SHIFT.
Signed-off-by: Jungseung Lee <js07.lee@samsung.com>
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
The driver calls le32_to_cpu() to convert the little-endian tables
to a CPU endianness, where le32_to_cpus() should have been called.
Was going to use that one... and then discovered a whole array
converter, le32_to_cpu_array()! :-)
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Pull MTD updates from Miquel Raynal:
"MTD core
- block2mtd: page index should use pgoff_t
- maps: physmap: minimal Runtime PM support
- maps: pcmciamtd: avoid possible sleep-in-atomic-context bugs
- concat: Fix a comment referring to an unknown symbol
Raw NAND:
- Macronix: Use match_string() helper
- Atmel: switch to using devm_fwnode_gpiod_get()
- Denali: rework the SKIP_BYTES feature and add reset controlling
- Brcmnand: set appropriate DMA mask
- Cadence: add unspecified HAS_IOMEM dependency
- Various cleanup.
Onenand:
- Rename Samsung and Omap2 drivers to avoid possible build warnings
- Enable compile testing
- Various build issues
- Kconfig cleanup
SPI-NAND:
- Support for Toshiba TC58CVG2S0HRAIJ
SPI-NOR:
- Add support for TB selection using SR bit 6,
- Add support for few flashes"
* tag 'mtd/for-5.6' of git://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux: (41 commits)
mtd: concat: Fix a comment referring to an unknown symbol
mtd: rawnand: add unspecified HAS_IOMEM dependency
mtd: block2mtd: page index should use pgoff_t
mtd: maps: physmap: Add minimal Runtime PM support
mtd: maps: pcmciamtd: fix possible sleep-in-atomic-context bugs in pcmciamtd_set_vpp()
mtd: onenand: Rename omap2 driver to avoid a build warning
mtd: onenand: Use a better name for samsung driver
mtd: rawnand: atmel: switch to using devm_fwnode_gpiod_get()
mtd: spinand: add support for Toshiba TC58CVG2S0HRAIJ
mtd: rawnand: macronix: Use match_string() helper to simplify the code
mtd: sharpslpart: Fix unsigned comparison to zero
mtd: onenand: Enable compile testing of OMAP and Samsung drivers
mtd: onenand: samsung: Fix printing format for size_t on 64-bit
mtd: onenand: samsung: Fix pointer cast -Wpointer-to-int-cast warnings on 64 bit
mtd: rawnand: denali: remove hard-coded DENALI_DEFAULT_OOB_SKIP_BYTES
mtd: rawnand: denali_dt: add reset controlling
dt-bindings: mtd: denali_dt: document reset property
mtd: rawnand: denali_dt: Add support for configuring SPARE_AREA_SKIP_BYTES
mtd: rawnand: denali_dt: error out if platform has no associated data
mtd: rawnand: brcmnand: Set appropriate DMA mask
...
Pull UBI/UBIFS updates from Miquel Raynal:
"This pull request contains mostly fixes for UBI and UBIFS:
UBI:
- Fixes for memory leaks in error paths
- Fix for an logic error in a fastmap selfcheck
UBIFS:
- Fix for FS_IOC_SETFLAGS related to fscrypt flag
- Support for FS_ENCRYPT_FL
- Fix for a dead lock in bulk-read mode"
Sent on behalf of Richard Weinberger who is traveling.
* tag 'upstream-5.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/ubifs:
ubi: Fix an error pointer dereference in error handling code
ubifs: Fix memory leak from c->sup_node
ubifs: Fix ino_t format warnings in orphan_delete()
ubifs: Fix deadlock in concurrent bulk-read and writepage
ubifs: Fix wrong memory allocation
ubi: Free the normal volumes in error paths of ubi_attach_mtd_dev()
ubi: Check the presence of volume before call ubi_fastmap_destroy_checkmap()
ubifs: Add support for FS_ENCRYPT_FL
ubifs: Fix FS_IOC_SETFLAGS unexpectedly clearing encrypt flag
ubi: wl: Remove set but not used variable 'prev_e'
ubi: fastmap: Fix inverted logic in seen selfcheck
Pull ioremap updates from Christoph Hellwig:
"Remove the ioremap_nocache API (plus wrappers) that are always
identical to ioremap"
* tag 'ioremap-5.6' of git://git.infradead.org/users/hch/ioremap:
remove ioremap_nocache and devm_ioremap_nocache
MIPS: define ioremap_nocache to ioremap
Raw NAND
* Macronix: Use match_string() helper
* Atmel: switch to using devm_fwnode_gpiod_get()
* Denali: rework the SKIP_BYTES feature and add reset controlling
* Brcmnand: set appropriate DMA mask
* Various cleanup.
Onenand drivers
* Rename Samsung and Omap2 drivers to avoid possible build warnings
* Enable compile testing
* Various build issues
* Kconfig cleanup
SPI-NAND
* Support for Toshiba TC58CVG2S0HRAIJ
Fix the comment describing what the mtd_concat_destroy() function
does. It referrers to the concat_mtd_devs symbol which has never
existed (at least not since the beginning of the Git era).
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Currently CONFIG_MTD_NAND_CADENCE implicitly depends on
CONFIG_HAS_IOMEM=y; consequently, on architectures without IOMEM we get
the following build error:
ld: drivers/mtd/nand/raw/cadence-nand-controller.o: in function `cadence_nand_dt_probe.cold.31':
drivers/mtd/nand/raw/cadence-nand-controller.c:2969: undefined reference to `devm_platform_ioremap_resource'
ld: drivers/mtd/nand/raw/cadence-nand-controller.c:2977: undefined reference to `devm_ioremap_resource'
Fix the build error by adding the unspecified dependency.
Reported-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Add minimal runtime PM support (enable on probe, disable on remove), to
ensure proper operation with a parent device that uses runtime PM.
This is needed on systems where the FLASH is connected to a bus
controller that is contained in a PM domain and/or has a gateable
functional clock. In such cases, before accessing any device connected
to the external bus, the PM domain must be powered up, and/or the
functional clock must be enabled, which is typically handled through
runtime PM by the bus controller driver.
An example of this is the Renesas APE6-EVM development board, which has
an Ethernet controller and a CFI FLASH connected to the Bus State
Controller (BSC) of an R-Mobile APE6 SoC.
As long as the Ethernet driver, which had Runtime PM support since
commit 3a611e26e9 ("net/smsc911x: Add minimal runtime PM
support"), keeps the BSC powered, accessing the FLASH works.
When the ethernet node in r8a73a4-ape6evm.dts is disabled, the BSC is
never powered up, and the kernel crashes when trying to access the
FLASH:
Unhandled fault: imprecise external abort (0x1406) at 0x00000000
pgd = (ptrval)
[00000000] *pgd=7fef2835
Internal error: : 1406 [#1] SMP ARM
CPU: 0 PID: 122 Comm: hd Tainted: G W 5.5.0-rc1-ape6evm-00814-g38ca966db25b9dbd-dirty #136
Hardware name: Generic R8A73A4 (Flattened Device Tree)
PC is at chip_ready+0x12c/0x380
LR is at chip_ready+0x10c/0x380
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
The driver may sleep while holding a spinlock.
The function call path (from bottom to top) in Linux 4.19 is:
drivers/pcmcia/pcmcia_resource.c, 312:
mutex_lock in pcmcia_fixup_vpp
drivers/mtd/maps/pcmciamtd.c, 309:
pcmcia_fixup_vpp in pcmciamtd_set_vpp
drivers/mtd/maps/pcmciamtd.c, 306:
_raw_spin_lock_irqsave in pcmciamtd_set_vpp
drivers/pcmcia/pcmcia_resource.c, 312:
mutex_lock in pcmcia_fixup_vpp
drivers/mtd/maps/pcmciamtd.c, 312:
pcmcia_fixup_vpp in pcmciamtd_set_vpp
drivers/mtd/maps/pcmciamtd.c, 306:
_raw_spin_lock_irqsave in pcmciamtd_set_vp
mutex_lock() may sleep at runtime.
To fix these bugs, the spinlock is replaced with a mutex.
These bugs are found by a static analysis tool STCheck written by
myself.
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
As previously reported by Sudip Mukherjee for the Samsung driver, the
omap2 onenand driver is called omap2.c in our directory and omap2.c in
the tty/serial/ directory. If both drivers are compiled as modules, it
would produce the following warning:
warning: same module names found:
drivers/tty/serial/omap2.ko
drivers/mtd/nand/onenand/omap2.ko
Rename the onenand omap2 driver so that it fits the folder's
convention: onenand_omap2.c.
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Commit 55ed51fff2 ("{tty: serial, nand: onenand}: samsung: rename to
fix build warning") has changed the samsung.c driver to be
samsung_mtd.c in order to avoid a conflict in module names with the
tty driver.
Since the *_mtd suffix is very undescriptive, rename it to
onenand_samsung.c, following the folder's convention. Same will be
applied to the omap2 onenand driver.
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
